(1) Field of the Invention
The present invention relates to a method for virtualizing a plurality of network storages and, more particularly, to a system and method for making a client computer virtually see a plurality of file systems as a single file system.
(2) Description of the Related Art
Hitherto, data is stored in a storage (DAS: Direct Attached Storage) directly connected to a client computer. Since data on a DAS can be accessed only via a client computer directly connected to the DAS, in the case where a client computer accesses data of a DAS connected to another client computer, the client computer has to access the data via the computer to which the DAS is connected.
However, because of development of network technologies of recent years, the amount of data to be stored in each client computer is rapidly increasing. In association with the increase, needs of efficient management of the data are increasing, and a storage system connected to a network (network storage) is being introduced in place of the DAS. By the introduction of the network storage, sharing of data among a plurality of client computers is enabled, and an administrator can efficiently manage shared data
As examples of the network storage, there are an SAN (Storage Area Network) storages connected to an SAN, for providing a block access, and an NAS (Network Attached Storage) for providing a file access, connected to the IP network, an infiniband, or the like. Among them, the NAS is expanding its market because it is cheaper than the SAN storage and by which a plurality of client computers can easily share a file by using a plurality of standard file access protocols such as NFS (Network File System) and CIFS (Common Internet File System).
As described above, the network storage is very popular since the administrator can efficiently manage shared data of a plurality of client computers in a lump. However, as a result of further increase in a shared data amount handled by a plurality of client computers as the network technologies develop, it is becoming difficult to store all of data by a single network storage. When the amount of data to be stored exceeds the capacity of a single network storage, the administrator has to prepare a new network storage and perform the following work.
First, the administrator connects the new network storage to the network and sets basical information such as an IP address and a name to the network storage. After that, the administrator has to set the sharing information for indicating which of the plurality of network storages should be shared by a plurality of client computers. For example, it is assumed that when a new network storage is added in a state where two client computers share a network storage, the administrator changes the setups so that a first client computer uses the existing network storage and a second client computer uses the new network storage. In the case of changing the sharing setting information among a plurality of client computers as described above, the administrator has to explicitly move the data of the second client computer on the existing network storage to the new network storage. The larger the number of network storages to be managed is, the higher the administration cost is.
As a method of suppressing the administration cost, a network storage virtualizing technique for virtualizing a plurality of network storages so as to be sensed by a client computer virtually as single storage system is proposed, in order to prevent an influence from being exerted on a whole system even when a new network storage is added.
For example, Japanese Patent Laid-open No. 2002-99519 (first prior art) discloses a method of constructing a dynamic file server system for virtualizing a plurality of file server systems connected to a network so as to be sensed by a client as a single dynamic virtual file server, in the case where the capacity of an existing file server is becoming short and is desired to be increased, without requiring expert knowledge of adding a disk to an existing file server and devices and work for extension. When a server computer transmits declaration information indicating that the server computer becomes a member of a dynamic virtual file server onto the network, each of the members of the dynamic virtual file server receives the declaration information, interprets the contents, and updates management information of the dynamic virtual file server. By adding the server system which has transmitted the declaration information as a new member of the dynamic virtual server, the storage capacity can be extended during operating the dynamic virtual file server.
Japanese Patent Laid-open No 2001-51890 (second prior art) discloses a virtual distributed file server system which does not make a client aware of the number of a plurality of file servers distributed on a network and a connection state of a storage. The virtual distributed file server system is distributed to a plurality of file servers connected to a network which can multicast. Each file server has a mapping table for holding mapping information between the virtual distributed file server system and a local file system for actually managing a file. Upon receiving a file access request multicasted from a client, each file server refers to the mapping table, and performs a file access operation on a local file system in the server only when the file server recognizes that the file server itself is the optimum one for processing the request. Since the client accesses a plurality of file servers distributed on the network via the virtual distributed file server, the system can be expanded without making the client aware of the existence of the plurality of file servers.
Japanese Patent Laid-open No. Hei 5-241934 (third prior art) discloses a virtual file system having the function of virtually showing a plurality of file systems as a single file system irrespective of an actual layout state. The virtual file system exists in one of a plurality of servers, manages the plurality of file systems in a single tree structure, and accepts file access requests from all of clients. A client issues a file access request to the server in which the virtual file system exists. When the file access request is received, the virtual file system investigates a file system of the server in which the requested file exits, issues an access request to the server in which the requested file exists, and transmits data obtained from the server to the client. It is sufficient for the clients and each of the members of the virtual file system to know only information of the virtual file system. Accordingly, a system change caused by adding a server can be easily realized.
However, the network storage virtualizing technique disclosed in the first, second, and third prior arts have been achieved on the premise that, in the case of adding a new network storage, existing network storages are already managed in a unified way by a virtualized-and-unified file system. In addition, all of members belonging to the virtualized-and-unified file system have to be provided with information regarding virtualization.
For example, in the first prior art, all of members belonging to the virtualized-and-unified file system have to possess information of members constructing the virtualized-and-unified file system. In the second prior art, all of members belonging to the virtualized-and-unified file system have to possess mapping information for correlating the virtualized-and-unified file system and a local file system. In the third prior art, all of members have to possess information of the server in which the virtualized-and-unified file system exists.
Generally, in the case of introducing a network storage, the administrator has not imagined that the capacity becomes short. Therefore, the administrator purchases one network storage, and starts operating it in an office or the like. When the capacity becomes short, the administrator purchases a new network storage. In a state where only one network storage operates, the virtualizing function is not necessary for the network storage in operation because it is unnecessary to virtually unify the network storage.
Only after a new network storage is purchased due to a capacity shortage, the administrator tries to manage the network storages in a unified manner in order to reduce administration cost. However, according to the known methods, since the existing network storage does not have the virtualizing function, the administrator cannot virtualize the network storages while maintaining a directories tree structure of the existing network storage. To solve the problem, in the case of adding a new network storage in a state where the existing network storage is operating without using the virtualizing function, it is necessary to realize virtualization while maintaining the directories tree structure of the existing network storage.
In most of the cases, the network storage is offered as an appliance. In order to reduce administration cost, in general, an appliance limits the setup information which can be changed by the administrator. Consequently, when the administrator tries to manage the existing network storage in a unified manner, which has been operated without virtualization functions, the administrator cannot set special information for virtualizing as described above. As a result, unified management cannot be realized. To solve the problem, in the case of managing an existing network storage and a new network storage in a virtually unified manner, it is required to realize the virtualization without setting special information into the existing network storage.
Further, in a state where the existing network storage is shared by a plurality of client computers, sharing setting information in the client computers have to be changed so as to be adapted to the new network storage, even if a new network storage additionally provided and an existing network storage can be virtually unified. To change the sharing setting information, the operation of each client computer has to be stopped once, so that administration cost increases. To solve the problem, the existing network storage and the new network storage have to be virtually unified without changing the sharing setting information of the network storages in the client computers, that is, while allowing the client computers to continuously share the existing network storage.